AMD Ryzen 5 2400G and Ryzen 3 2200G APUs performance unveiled

[french toast]

YET, amd avoided this kind of comparisons now and decided to only target Intel IGP with is nothing.

Fair point, but still it should not be expected that they compare Apu Vs dgpu.
Intel don't.

Those examples were shown because they show AMD products in a good light, it's obvious on standard or even fairly high clocked memory that it is not the case this time..or it is too close for comfort, either way it's not a good strategy.
Saying that amd did show some slides at Ryzen launch painting Ryzen In a bad light Vs broadwell E in gaming.

 

[Shivansps]

Isn't it something like this: ~250MB is reserved for the GPU and it sees that portion as VRAM, if more is needed it'll start using the "main system memory"? like if a dGPU runs out of VRAM and starts dumping less used data to RAM. Only in this case the VRAM and system RAM are physically the same.

It seems to say very little ram is being used for the graphics, less than 250mb, and 6.5gb of ram is being used for the game?

Rivatuner does not report the correct amount of memory, it will only report the fixed memory, GPU-Z in other hand reports correctly the amount used for both fixed and dynamic for APU.

Also games tend to use all VRAM avalible regardless, that "6.5GB" is the game fighting the gpu driver for memory, even worse, Windows could be sending some of the dynamic VRAM to the pagefile on that example. I cant really be sure since i dont know how much of the dynamic memory is used.

 

[neblogai]

The difference should be quite diminishing between the two, since increasing the GPU frequency by 24% only resulted in 8.2% difference in performance (at 3200MHz MEMCLK).
The difference between the 2200G and 2400G SKU is 56%, but I don't expect the actual difference to be higher than < 15% with everything else being equal.

I wonder, can enough power be supplied for overclocking the iGPU- and if there is throttling, even here in 3DMark. Previous APUs did seem to be very limited in how much total power can be used by them, and would lower CPU clocks when GPU is loaded. Raven Ridge is also mobile oriented, with all it's parts aiming to be most efficient at low power consumption. I wonder if it's power delivery will be enough for overclocking on desktop where CPU is at near 4GHz, and iGPU highly overclocked. These 3DMark tests do not put load on both CPU and GPU at the same time like games do- and still- OCing of iGPU does not show much benefit. I hope it is a result of limited memory bandwidth, and is not caused by power limit.

 

[The Stilt]

I wonder, can enough power be supplied for overclocking the iGPU- and if there is throttling, even here in 3DMark. Previous APUs did seem to be very limited in how much total power can be used by them, and would lower CPU clocks when GPU is loaded. Raven Ridge is also mobile oriented, with all it's parts aiming to be most efficient at low power consumption. I wonder if it's power delivery will be enough for overclocking on desktop where CPU is at near 4GHz, and iGPU highly overclocked. These 3DMark tests do not put load on both CPU and GPU at the same time like games do- and still- OCing of iGPU does not show much benefit. I hope it is a result of limited memory bandwidth, and is not caused by power limit.

Overclocking the APU (i.e. OC-Mode) should disable all power limiters on Raven, same way it does on Zeppelin and Pinnacle.
The power consumption of the iGPU could reach levels (at high overclocks) which cause the weak (usually 2-phase) VRM to overheat, but that's another thing.

 

[whm1974]

OK folks I've notice some of us are starting to get carried away here, so let us not over hype these products before reviewers can properly test them. And yes I'm rooting for AMD here and I believe that Raven Ridge will be very competitive with Intel's offerings. Both APUs do seem well suited for budget gamers, and should fly off the shelves for this market segment.

 

[CatMerc]

I just noticed something interesting with AMD 25x20 graph on this article: https://www.anandtech.com/show/1231...n-exclusive-interview-with-dr-lisa-su-amd-ceo

Spoiler

We're dealing with exponents here.

Kaveri is 2014 base at 1x
In order to reach 25x efficiency by 2020, efficiency needs to increase by 1.71x every year (6√25) from the previous product.

Therefore, the targets for each year are:
2014: 1x
2015: 1.71x
2016: 2.92x
2017: 5x
2018: 8.55x
2019: 14.62x
2020: 25x

AMD is well ahead right now of their targets thanks to the headroom afforded by Carrizo, but they need another major jump or risk falling behind. Going from Bristol to Raven they gained 1.47x, which is below the 1.71x yearly target, and that was with going from 28nm to 14nm, and from construction cores to Zen. To not fall behind in 2018, Picasso (Raven Refresh), would need to be 46% more efficient than Raven Ridge. A tall order of a refresh.

 

[vinhom]

@CatMerc You're absolutely correct. I'm sorry for bringing this up.

 

[Shivansps]

Fair point, but still it should not be expected that they compare Apu Vs dgpu.
Intel don't.

Those examples were shown because they show AMD products in a good light, it's obvious on standard or even fairly high clocked memory that it is not the case this time..or it is too close for comfort, either way it's not a good strategy.
Saying that amd did show some slides at Ryzen launch painting Ryzen In a bad light Vs broadwell E in gaming.

You need to realise that Intel IGP were never good, there were a few cases, specially on mobile, where they yielded a acceptable result, same happened on a few cases con desktop (AM1 vs Haswell Celeron/Pentium IGP), Skylake Celeron/Pentium vs A4-7300. But thats petty much it.

AMD IGP were good vs Intel ever since they showed up, like AMD 690G/780G/880G. With APUs AMD started to target entry level dGPUs because the IGP battle was long won, with the old gen AMD always tried to compare APU vs entry level dGPUs. This comparison vs what is already a 3 gens old Intel IGP is not really interesting. At least not to me. And it feels like a regresion rather than a step forward.

We are close to see their true performance, so will soon see. But i can tell you already, if 2400G cant match a GT1030 with fast rams and NO OVERCLOCK this is already a big step back compared to older AMD APU gens. Beating Intel IGP is old news at this point.

 

[whm1974]

You need to realise that Intel IGP were never good, there were a few cases, specially on mobile, where they yielded a acceptable result, same happened on a few cases con desktop (AM1 vs Haswell Celeron/Pentium IGP), Skylake Celeron/Pentium vs A4-7300. But thats petty much it.

AMD IGP were good vs Intel ever since they showed up, like AMD 690G/780G/880G. With APUs AMD started to target entry level dGPUs because the IGP battle was long won, with the old gen AMD always tried to compare APU vs entry level dGPUs. This comparison vs what is already a 3 gens old Intel IGP is not really interesting. At least not to me. And it feels like a regresion rather than a step forward.

We are close to see their true performance, so will soon see. But i can tell you already, if 2400G cant match a GT1030 with fast rams and NO OVERCLOCK this is already a big step back compared to older AMD APU gens. Beating Intel IGP is old news at this point.

I don't see how this would a big step back if RR outperforms previous APUs in both iGPU and CPU performance by a large margin. I'm sure current owners of previous APUs will be happy with upgrading to the 2200G and 2400G APUs once out.

 

[piesquared]

I don't see how this would a big step back if RR outperforms previous APUs in both iGPU and CPU performance by a large margin. I'm sure current owners of previous APUs will be happy with upgrading to the 2200G and 2400G APUs once out.

It's not a step back in any way, it's a big step forward actually. Not only because the onboard Vega graphics can match a discrete GPU like a 1030, but because it is integrated with a CPU on die that has very good performance and can compete with the newest CPUs on the market. This is something the market has never seen before. It's very simple and inexpensive for OEMs to build their products and therefore can offer lower cost products to consumers, and/or increase their margins. It will probably be a combination of the 2, and either way, OEMs will be eager to adopt the platform. It will be even more attractive to them since it is AM4 socket compatible. AM1 was quite popular in lower income countries for DiY builders, this will be even more popular.

 

[Glo.]

We're dealing with exponents here.

Kaveri is 2014 base at 1x
In order to reach 25x efficiency by 2020, efficiency needs to increase by 1.71x every year (6√25) from the previous product.

Therefore, the targets for each year are:
2014: 1x
2015: 1.71x
2016: 2.92x
2017: 5x
2018: 8.55x
2019: 14.62x
2020: 25x

AMD is well ahead right now of their targets thanks to the headroom afforded by Carrizo, but they need another major jump or risk falling behind. Going from Bristol to Raven they gained 1.47x, which is below the 1.71x yearly target, and that was with going from 28nm to 14nm, and from construction cores to Zen. To not fall behind in 2018, Picasso (Raven Refresh), would need to be 46% more efficient than Raven Ridge. A tall order of a refresh.

Only thing that can to mind for 2018, and maintaining the growth of efficiency is HBM2 APU design.

 

[CatMerc]

Made a chart to illustrate what I mean:

The 12nm lines assume that Raven Ridge refresh will get a 15% perf/watt improvement based purely on GloFo's 12nm characteristics. It might be more, might be less. After that it assumes with Zen2 and beyond AMD follows the trend exactly.

 

[whm1974]

It's not a step back in any way, it's a big step forward actually. Not only because the onboard Vega graphics can match a discrete GPU like a 1030, but because it is integrated with a CPU on die that has very good performance and can compete with the newest CPUs on the market. This is something the market has never seen before. It's very simple and inexpensive for OEMs to build their products and therefore can offer lower cost products to consumers, and/or increase their margins. It will probably be a combination of the 2, and either way, OEMs will be eager to adopt the platform. It will be even more attractive to them since it is AM4 socket compatible. AM1 was quite popular in lower income countries for DiY builders, this will be even more popular.

If I was a betting man, I would bet that both AM4 APUs will sell out causing Intel to start panicking really bad.

 

[Shivansps]

I don't see how this would a big step back if RR outperforms previous APUs in both iGPU and CPU performance by a large margin. I'm sure current owners of previous APUs will be happy with upgrading to the 2200G and 2400G APUs once out.

Because they lagged behind dGPUs. What would have happen to old gen APU if they were like that?

 

[LTC8K6]

I doubt you can run the CPU and GPU wide open at the same time.

Maybe you can with more efficient chips, but we will have to see if throttling rears it's ugly head or not.

 

[whm1974]

Because they lagged behind dGPUs. What would have happen to old gen APU if they were like that?

Everything else being equal, and aside from the very low end, iGPUs will always be lagging behind dGPUs. What did you expect to happend?

 

[xblax]

If I was a betting man, I would bet that both AM4 APUs will sell out causing Intel to start panicking really bad.

That would then drive prices up above msrp ... I really hope AMD can meet the demand. Assuming that Epyc could be ramping up and sell in high volume and Zen+ sells good AMD could get in bigger troubles with Glofo's capacity.

But another thing came to my mind. What mainboards should people buy at launch? If the current mainboards cannot boot to bios without updates many people could run into troubles ... it would make sense to launch 400-series boards together with the APUs. On the other hand the selection of boards is quite bad for APUs. Out of 75 boards that have APU support only 10 have a Display Port. Some boards even have a single DVI port, wtf?

 

[whm1974]

That would then drive prices up above msrp ... I really hope AMD can meet the demand. Assuming that Epyc could be ramping up and sell in high volume and Zen+ sells good AMD could get in bigger troubles with Glofo's capacity.

But another thing came to my mind. What mainboards should people buy at launch? If the current mainboards cannot boot to bios without updates many people could run into troubles ... it would make sense to launch 400-series boards together with the APUs. On the other hand the selection of boards is quite bad for APUs. Out of 75 boards that have APU support only 10 have a Display Port. Some boards even have a single DVI port, wtf?

That is an issue I have with both AMD and Intel boards. Seriously both HDMI and Displayport should be available, with no VGA or DVI ports anywhere to be seen.

 

[Abwx]

The 12nm lines assume that Raven Ridge refresh will get a 15% perf/watt improvement based purely on GloFo's 12nm characteristics. It might be more, might be less.

It should be twice this value, something like 30% improvement at iso frequency, based precisely on GF s 12nm chracteristics...

 

[Shivansps]

Everything else being equal, and aside from the very low end, iGPUs will always be lagging behind dGPUs. What did you expect to happend?

Thats not true, older gen APU allowed to get better performance than basic low end gpus, AMD itselft promoted that, ive already showed you. Thats was the whole point in an APU. Better than R7 250 DDR3, better than GT740 DDR3 on Kaveri... and AMD promoted that to no end. Now the lowest is the GT1030, and with 64 bit GDDR5 this is like the older DDR3 gpus, it can be beaten with DDR4.

Ill say this clear, if 2400G cant beat, or at the very least match a GT1030, this will be the first time since the APUs came to market that they are unable to match an entry level gpu.

And dont get me wrong, i want the 2400G to beat a GT1030, and i want the 2200G to be about 10% behind, i just dont see this as possible, and to me this is bad.

 

[frozentundra123456]

Fair point, but still it should not be expected that they compare Apu Vs dgpu.
Intel don't.

Those examples were shown because they show AMD products in a good light, it's obvious on standard or even fairly high clocked memory that it is not the case this time..or it is too close for comfort, either way it's not a good strategy.
Saying that amd did show some slides at Ryzen launch painting Ryzen In a bad light Vs broadwell E in gaming.

For desktop, a dgpu is the alternative option to an APU, so it certainly is a comparison the end user wants to make. Perhaps not for AMD or their supporters, because it shows just how weak an APU really is for gaming.

 

[prtskg]

Thats not true, older gen APU allowed to get better performance than basic low end gpus, AMD itselft promoted that, ive already showed you. Thats was the whole point in an APU. Better than R7 250 DDR3, better than GT740 DDR3 on Kaveri... and AMD promoted that to no end. Now the lowest is the GT1030, and with 64 bit GDDR5 this is like the older DDR3 gpus, it can be beaten with DDR4.

Ill say this clear, if 2400G cant beat, or at the very least match a GT1030, this will be the first time since the APUs came to market that they are unable to match an entry level gpu.

And dont get me wrong, i want the 2400G to beat a GT1030, and i want the 2200G to be about 10% behind, i just dont see this as possible, and to me this is bad.

There was R7 250 or GT740 with ddr3 because there was market for it. Now integrated gpu has eaten that market. Hence the minimum is GT1030 now, with GDDR5. It must be pointed out because some people don't see it that these low level gpus no longer come with ddr but gddr. The reason is obvious, ddr gpus aren't competitive now. There is a huge market for apu as in this world of 7 Billion people, less than 10 million would be playing Witcher 3 now. Let's not forget those 6.9B for 10M.

For desktop, a dgpu is the alternative option to an APU, so it certainly is a comparison the end user wants to make. Perhaps not for AMD or their supporters, because it shows just how weak an APU really is for gaming.

Gaming at FHD with AAA games is what integrated graphics are weak at. But how many do that? There is a reason integrated graphics chips are called volume products. Some people are unfortunately not seeing the forest for the tree.

 

[french toast]

For desktop, a dgpu is the alternative option to an APU, so it certainly is a comparison the end user wants to make. Perhaps not for AMD or their supporters, because it shows just how weak an APU really is for gaming.

Is this trolling? These Apu are not designed as dedicated gaming devices, they just have competent graphics when compared to competitor's products, they are decent enough that if you equip with decent memory and overclock (with standard wraith cooler) you can comfortably play AAA games at low/medium settings...something you can't do with intel products...without buying a dedicated GPU.

You and others arguements/remarks make no sense, AMD does not restrict their Apu to IGP graphics..YOU CAN STILL CONNECT ANY DGPU YOU LIKE...
Gaming..(sub £200)
You put forward a i3-8100..I counter with 2200G/2400G
You counter with i3-8100 + gtx 1030..or even RX 550..
I say R3 1200 + Rx 560 or gtx 1050.
The only combo that can't be beaten for sub £200..is g4560+ 1050Ti.
The closest AMD can get with current products is R3 1200 + Rx 560 4gb.
..which loses a close FPS/£ battle... but offers better all round performance and value long term.

Back on topic, APU Vs APU AMD is the market leader, it offers products with the best all round balance, where you can buy one processor and have a decent experience in productivity and light gaming without breaking the bank...with the option to upgrade in the future or add dgpu if required.
That is picking one specific scenario, ..and not knowing what will happen to R3 1200 prices when 2200g hits...they might well get close to a pentium..
I don't see what your point is, there are no downsides here.

 

[krumme]

For desktop, a dgpu is the alternative option to an APU, so it certainly is a comparison the end user wants to make. Perhaps not for AMD or their supporters, because it shows just how weak an APU really is for gaming.

Dont dismiss it so early. Its also a good choice for old, bitter men with low income.

 

[AtenRa]

Thats not true, older gen APU allowed to get better performance than basic low end gpus, AMD itselft promoted that, ive already showed you. Thats was the whole point in an APU. Better than R7 250 DDR3, better than GT740 DDR3 on Kaveri... and AMD promoted that to no end. Now the lowest is the GT1030, and with 64 bit GDDR5 this is like the older DDR3 gpus, it can be beaten with DDR4.

Ill say this clear, if 2400G cant beat, or at the very least match a GT1030, this will be the first time since the APUs came to market that they are unable to match an entry level gpu.

And dont get me wrong, i want the 2400G to beat a GT1030, and i want the 2200G to be about 10% behind, i just dont see this as possible, and to me this is bad.

You keep talking about how those Ryzen APUs will not reach GT1030 performance but you still havent provided any technical reasons as to why you believe this will happen. What technically make you feel that 2400G with 3200MHz memory will not come close to GT1030 ??

GT730 with 64bit GDDR-5 was neck and neck with A10-7870K paired with DDR-3 2133MHz, why do you believe that GT1030 will be that much faster than 2400G + DDR-4 3200 ??


Edit: Also for those that compare apples to oranges, 2200G at 99$ is a direct competitor to Core i3 8100 ($117) and 2400G at $169 is a direct competitor to Core i3 8350K ($169) and Core i5 8400 ($182).